Business copy issue

[zxcv]

Hi ,

We have eva6400 and 16 big clients with oracle 9i database.

Our daily backup strategy is to take a snapclone of this clients.

the question i have is ,should i wait for the levelling to complete in storage or i can take export backup immediately after RSM script is completed.

[hvhari]

You can export backup immediately after RSM script is completed. Levelling is an internal "low priority" process within EVA and it does not affect redundancy of the data.

[zxcv]

Hi Hari ,

thanks for ur reply.

But a snapclone unless the levelling is completed , if i fire a export it will sources the original vdisk right ??

[hvhari]

"Snapclone normalization (unsharing)" is the process which verifies the new snapclone vdisk with original one. Till this process is complete data may be sourced from original vdisk.

Snapclone normalization (unsharing)

The snapclone normalization is a process which verifies that data is identical on a source virtual

disk and its snapclone, before the snapclone can become an independent disk (a point-in-time

copy). See

Snapclones. Snapclone normalization is also called unsharing.

Some actions should not be performed during snapclone normalization. When creating jobs, use

wait commands to ensure that normalization is completed. See the

WaitStorageVolumeNormalization, WaitStorageVolumesNormalization, and

WaitVolumeGroupNormalization job commands.

The guide http://bizsupport1.austin.hp.com/bc/docs/support/SupportManual/c02838357/c02838357.pdf has the more details on this.

[zxcv]

Hi Hari ,

We have observed that when we fire a snapclone & mount it on a backup server , and dont wait for levelling to finish, the export  backup just gets choked and in turn hampers my main jobs running on Source Vdisk.

 As per that document during unsharing we can read-write, but serious performance issue is observed on production.

Instead if we wait for levelling for around 30 min everything goes well.

but our backup window is getting smaller and smaller the data/clients are increasing rapidly.

How can WaitVolumeGroupNormalization job help us in this case ?

[hvhari]

These wait commands essentially adds time to the job before its marked complete. When normalisation is complete, the snapclone becomes an independent vdisk and further read/writes are send from this new copy (hence the source vdisk is not given additional workload). The backup for this vdisk can happen irrespective of the status of source vdisk, hence ideally you should not be worrying about backup window time. (ie. you can even backup this new vdisk during peak timings where source vdisk serves I/Os and the new vdisk/snapclone serves backup).

[zxcv]

Hi Hari ,

> you can even backup this new vdisk during peak timings where source vdisk serves I/Os and the new vdisk/snapclone serves backup)

but thats not happening.

When we are trying to take backup ( export )  it simply chokes both the new vdisk as well as source vdisk.

[hvhari]

After "normalisation" is completed there should not be any effect/reference to source vdisk. If this is happening, will need to check exactly what kind of IO is coming to vdisk.

May be HP Support can validate the data on performance parameters (once you collect them via EVAPerf tool when you have issues).

You may also wait to see if any one else on this forum has come across this situation earlier.